A retrospective case-series study evaluated the prognosis of 853 dogs with acute kidney injury (AKI) based on the RIFLE (Risk, Injury, Failure, Loss and End-stage renal failure) criteria, derived from human medicine. The 30-day mortality of dogs with AKI in each class was found to be 23.8 per cent (40 of 168) dogs for Risk, 41.0 per cent (107 of 261) dogs for Injury and 78.5 per cent (333 of 424) dogs for Failure. Using the dogs in the Risk class as the reference, the mortality of dogs in either the Injury or Failure class was significantly higher than that of dogs in the Risk class (P<0.05). The longest median survival time was observed in the Risk class (nine days) and the shortest median survival time was observed in the Failure class (three days). Using a multiple logistic regression model, a new score that simultaneously considered RIFLE class, diarrhoea status and serum phosphorus level was calculated to predict prognosis. Evaluation using the area under the receiver-operating characteristic curve (AUROC) indicated that the new scoring method (AUROC 0.80) was a better prognostic indicator than using RIFLE criteria alone (AUROC 0.73).
Statistics from Altmetric.com
ACUTE kidney injury (AKI) is a complex and common clinical disorder in both human and veterinary medicine. Patients with AKI show differing clinical manifestations that range from a slight increase in serum creatinine to severe renal failure that characteristically portends high mortality. It is therefore important to determine prognosis following AKI in epidemiological studies. However, because of the lack of a universal standard classification for AKI, the severity of AKI remains controversial. In human medicine, mortality due to AKI has been reported to vary widely between studies (Liaño and Pascual 1996, Silvester and others 2001, Bagshaw and others 2005), and the results of similar studies in veterinary medicine have led to differing conclusions (Behrend and others 1996, Vaden and others 1997). Recently, a study in human beings was conducted to establish a uniform standard for AKI diagnosis and classification in order to allow better management of patients. A multi-level classification system was developed by the Acute Dialysis Quality Initiative (ADQI) working group (Bellomo and others 2004). The system is known as RIFLE (Bellomo and others 2004), with the acronym indicating Risk of renal dysfunction, Injury to the kidney, Failure of renal function, Loss of renal function and End-stage renal disease (Table 1). This multi-level system simultaneously classified the renal function of patients into three classes of severity: those in whom renal function was mildly affected (Risk class), those in whom renal dysfunction was marked (Injury class), and those in whom renal damage was severe (Failure class). Both sensitivity and specificity were considered when evaluating a wide range of renal diseases. RIFLE criteria were simple and easy to apply to clinical cases. To date, several studies have shown the validity of using RIFLE criteria when classifying AKI with regard to both severity and outcome (Hoste and others 2006, Uchino and others 2006, Bagshaw and others 2008, Ricci and others 2008).
RIFLE classification was based on relative changes to glomerular filtration rate (GFR) or urine output (Bellomo and others 2004). As the volume of urine output for 24 hours is not easily obtainable for each patient, an increase in the serum creatinine level was applied to evaluate GFR in human studies (Ricci and others 2008). However, as a premorbid baseline value for serum creatinine concentration is not available, the ADQI working group recommended a way of estimating a value for baseline serum creatinine based on the modification of diet in renal disease equation (Perez Valdivieso and others 2008, Ricci and others 2008). This method used a lower limit for normal GFR of 75 ml/min per 1.73 m2. The simplified formula showed the relationship between GFR and baseline serum creatinine based on age, race and sex (Levey and others 1999). In dogs, previous studies have also developed a simplified model showing the relationship between GFR and creatinine level (Finco and others 1995, Haller and others 1998).
At present, similar consensus criteria have not been established for dogs or cats in small animal practice. In fact, veterinarians also need a standard classification to evaluate acute renal disease for both epidemiological and clinical purposes. In this study, the authors postulated that the AKI criteria used in human beings could be applied in veterinary medicine, and that the prognosis of AKI could be determined not only by RIFLE criteria but also by other clinical factors. Use of a multiple logistic regression model, incorporating several clinical factors, could be helpful for prognostic prediction because the method has been successfully applied in diseases with multifactorial causes (Jones and others 1999, Skrifvars and others 2003). Therefore, the objectives of this study were to evaluate the correlation between RIFLE criteria (for Risk, Injury and Failure) and mortality of dogs with AKI, as well as to evaluate prognosis using these criteria.
Materials and methods
All dogs diagnosed with AKI (blood urea nitrogen [BUN] >10 mmol/l, creatinine >116 µmol/l, without previous renal disease) at the Veterinary Medical Teaching Hospital, National Chung Hsing University, Taiwan, between January 2000 and December 2006, were included in the study. Dogs with chronic renal disease (a clinical history of more than 14 days) with urinary obstruction or with bladder rupture were excluded from the study. Outcomes for the Loss of renal function and End-stage renal disease RIFLE classes were not studied.
Data collection and definition
The clinical information recorded included: demographic information, including age, sex and bodyweight; clinicopathological data, including values for haemoglobin, packed cell volume, white blood cell (WBC) count, serum creatinine, BUN, aspartate transaminase, alanine aminotransferase, lactate dehydrogenase, creatine kinase, albumin, calcium, phosphorus, glucose, serum potassium, sodium and chloride; clinical signs, such as vomiting, diarrhoea, anorexia and urinary abnormalities (polyuria, pollakiuria, oliguria and dysuria); and the duration in hospital. According to a previous study in human beings (Bellomo and others 2004), there are two alternative ways to determine RIFLE classification: by calculating either the relative increase in the creatinine level from baseline, or the percentage decrease in GFR (Table 1). Finco and others (1995) showed that GFR in dogs was equal to 227 × (1/serum creatinine concentration) (recalculated from the original data to accommodate SI units: P-creatinine [µmol/l], GFR [ml/min/kg]), with a 95 per cent confidence interval (CI) of ±0.5 ml/min/kg. This equation was used in the present study to estimate the relative standard levels of serum creatinine to determine the RIFLE category for the dogs with AKI. As the reference range for GFR was 2.61 to 4.1 ml/min/kg (Finco and others 1981, van den Brom and Biewenga 1981, Krawiec and others 1986, Heiene and Moe 1998), in order to prevent overestimation of renal function, the lowest endogenous GFR level (2.61 ml/min/kg) was used as the baseline value for GFR in dogs. This GFR baseline value was then used to calculate the range of GFR values for each RIFLE category, in terms of a 25, 50 and 75 per cent decrease in GFR. Accordingly, relative standard levels of serum creatinine were estimated using the equation developed by Finco and others (1995) for the different GFR values of each RIFLE category. The reference ranges for serum creatinine in each RIFLE class were estimated to be 116 to 174 µmol/l for Risk, 175 to 347.9 µmol/l for Injury and >347.9 µmol/l for Failure (Table 1). Finally, 30-day mortality was estimated for dogs with AKI for each RIFLE category.
Analysis was performed using SPSS v16.0 for Windows (SPSS). Normally distributed variables were reported as mean (sd) and analysis of variance (ANOVA) was applied. For continuous data, such as age and bodyweight, which approximately followed a normal distribution, one-way ANOVA with F-statistics was applied to compare more than two group means, with the mean (sd) for each variable being reported. For laboratory results that were not normally distributed, the median and interquartile range were reported, and the Kruskal-Wallis test was applied for statistical analysis. Categorical data were analysed by the chi-squared test or Fisher's exact test to evaluate any significant differences among groups. The overall 30-day survival rate in dogs with AKI in different RIFLE categories was analysed by Kaplan-Meier methods; the differences between categories were statistically evaluated by the log-rank test. To develop a new index for prognostic prediction, univariate analysis by logistic regression was applied first to identify any demographic and clinical factors possibly associated with death due to AKI. Odds ratios and their 95 per cent CIs were calculated. Multiple logistic regression analysis was then used to establish a prediction model. Furthermore, a stepwise analysis was applied to select the most significant predictors of mortality. The log (odds) value, based on the multiple logistic regression model, was then used as a new score for dogs with AKI when considering their RIFLE classification and relevant demographic and clinical information. Finally, when predicting prognosis, the area under the receiver-operating characteristic curve (AUROC) was used to compare performance when using RIFLE criteria alone and when using the new score calculated by the log (odds) based on the multiple logistic regression model. All statistical tests took P<0.05 to be significant.
Of the 853 dogs presented with AKI, the nine breeds most frequently represented were Maltese (12 per cent), pomeranian (9 per cent), poodle (5 per cent), shih tzu (5 per cent), schnauzer (5 per cent), Yorkshire terrier (4 per cent), cocker spaniel (3 per cent), chihuahua (2 per cent) and German shepherd dog (2 per cent); 40 per cent of the dogs were crossbred. A total of 168 cases were classified as Risk, 261 cases as Injury and 424 cases as Failure. The 30-day mortality in the Risk class was 23.8 per cent, which was significantly lower than that in the Injury (41.0 per cent) and Failure (78.5 per cent) classes (P<0.05). The distribution in terms of age, sex and bodyweight of dogs with AKI was not significantly different among the three RIFLE classes. However, the percentage of dogs with AKI showing signs of vomiting increased from 48.1 per cent in the Risk class to 60.5 per cent in the Injury class and 76.3 per cent in the Failure class (Table 2).
Values measured for WBC count, BUN, creatinine, albumin, total protein, phosphorus, chloride, glucose and potassium differed significantly between the three RIFLE classes. With the exception of WBC count and chloride, dogs in the Failure class had the highest clinical values for all laboratory examination results. The lowest value for chloride was observed in the Failure class, while the highest value for WBC count was observed in the Injury class (Table 3).
Factors associated with mortality, evaluated by univariate analysis, are shown in Table 4. Using the Risk class as the reference, odds ratios of death were 2.65 (95 per cent CI 1.67 to 4.19) for the Injury class and 14.68 (95 per cent CI 9.22 to 23.39) for the Failure class. Dogs showing clinical signs of vomiting and diarrhoea also had a significantly higher odds of death than dogs without these clinical signs. Dogs with higher levels of serum potassium and phosphorus, and a lower level of serum chloride, were found to be more likely to die due to AKI (Table 4). Following multiple logistic regression analysis, the RIFLE class, serum phosphorus concentration and clinical signs of diarrhoea were found to be the most significant predictors of death due to AKI. A new score based on the logistic linear model was constructed:
New score = log (odds) = 0.437P + 0.669D + 0.373C1 + 1.199C2
where P represents the phosphorus concentration (mmol/l); D is equal to 1 if the dog is suffering from diarrhoea, or 0 without diarrhoea; and the C1, C2 sets are (0, 0), (1, 0) and (0, 1) for the classes Risk, Injury and Failure, respectively. Therefore, a new score based on this logistic linear function was calculated for each dog with AKI. The AUROC was then calculated to evaluate the ability of the new scores to predict the survival rate. On the basis of the new score, the AUROC was 0.80 and the cut-off value was 2.85, with a sensitivity of 68.9 per cent and specificity of 80.7 per cent (Fig 1). This contrasts with the use of RIFLE criteria alone, where the AUROC was 0.73. For RIFLE criteria alone, using the Failure class to determine the survival rate, the sensitivity was 69.4 per cent and the specificity was 78 per cent. As shown in Fig 2, the cumulative 30-day survival rates for dogs with AKI in the three RIFLE classes differed significantly by the log-rank test (P<0.05). Median survival time decreased from nine days for dogs in the Risk class to five days in the Injury class and to three days in the Failure class (Fig 2).
The results of the present study indicated that, as in human medicine (Hoste and others 2006, Uchino and others 2006, Bagshaw and others 2008, Ricci and others 2008), RIFLE is a promising classification for dogs with AKI. The feasibility of RIFLE-like criteria for classification and as a prognostic indicator for dogs with AKI was highlighted in several ways. First, 30-day mortality was shown to increase with the severity of RIFLE class. Secondly, a mathematical model was developed that used not only RIFLE criteria but also included other clinical information for predicting survival. The results suggested that when simultaneously considering the phosphorus level and diarrhoea status of the dog together with RIFLE criteria, the results might have a better prognostic value than considering RIFLE criteria alone.
RIFLE criteria are based on serum creatinine, the analyte most frequently measured in small animal clinical chemistry, and used as an indirect measurement of GFR (Braun and others 2003). In human beings, it has been shown that the relationship between GFR and creatinine can be influenced by race, age and sex (Bellomo and others 2004). The relationship between GFR and creatinine reported by Finco and others (1995), without consideration of other factors, was applied to develop the RIFLE criteria in dogs with AKI in the present study. This method was used for the following reasons: first, in veterinary medicine, previous studies have found that sex does not appear to influence the relationship between GFR and serum creatinine in dogs (Finco and others 1995). Secondly, another study also indicated that there was no difference in serum creatinine concentration between young (less than one year) and older (more than nine years) dogs (Vajdovich and others 1997). Finally, it is believed that breed may influence the relationship between GFR and creatinine, and that hounds (especially greyhounds) have been reported to have higher creatinine levels (Feeman and others 2003) but with the same GFR (Drost and others 2006). In the present study, no hounds were included in the study population; therefore, the relationship between GFR and creatinine derived by Finco and others (1995) should still be appropriate for this analysis.
The three RIFLE groups showed no differences in terms of age, sex and bodyweight, or for various other clinical signs, such as urinary abnormalities, diarrhoea and anorexia. These findings demonstrated that the classes of renal injury were not affected by these factors and different grades of renal injury could be present with similar clinical signs. Although urinary abnormalities are important in AKI, and urine output is the other factor that can be used to determine RIFLE criteria (Bellomo and others 2004), no significant difference in urinary abnormalities was present among the three RIFLE classes in the present study. However, this might be due to a lack of standard, and descriptions of the urinary abnormalities can be subjective. Further studies are required to verify the real urine output in dogs with AKI. Vomiting is the most common clinical sign in dogs with uraemia (Peters and others 2005). The findings of the present study indicated that it is important to note that in the RIFLE classification of dogs with AKI, those in a more severe class had a greater chance of showing signs of vomiting. Therefore, vomiting appeared to indicate a more advanced grade of renal injury.
In the present study, the relationship between various clinicopathological factors and RIFLE criteria was also evaluated. Haematological analysis revealed that the WBC count in the Injury class appeared to be higher than that in other groups. However, this did not correlate with the severity of renal disease as estimated by the elevation of serum creatinine. This result might imply that WBC count may be deterministic only during the early stages of renal injury. Although the glucose, albumin and total protein levels increased with the severity of renal injury, these factors were not significantly associated with mortality in the present study. Such findings are reasonable, as these three factors are easily adjusted by fluid therapy, before the animals progress to a more severe phase of disease.
With regard to prognosis for AKI, higher levels of serum creatinine and phosphorus, a lower level of serum calcium and older age have been reported to contribute to the death of uraemic dogs (Behrend and others 1996, Vaden and others 1997). In the present study, as well as RIFLE criteria (serum creatinine as the standard), dogs showing signs of vomiting or diarrhoea, higher levels of BUN, phosphorus or potassium, and a lower level of chloride, had an increased odds of mortality. In contrast to a previous investigation by Vaden and others (1997), the present study indicated that older age or a lower level of serum calcium was not associated with higher mortality. This discrepancy may be due to different sample sizes (99 dogs included in the previous study v 853 in the present study) or different causes of renal damage in different regions. However, the detailed reasons for this phenomenon are unclear and further larger-scale epidemiological studies are needed. Nevertheless, dogs that presented with gastrointestinal clinical signs (vomiting or diarrhoea) or with aberrant electrolyte metabolism (high phosphorus and potassium, and low chloride levels) consistently showed a poor prognosis.
When used as a classification system in a human clinical setting, the RIFLE criteria do have some shortcomings. It should be noted that cases of prerenal azotaemia cannot be excluded using the RIFLE criteria, and different causes of renal injury or dysfunctions of other systems are neither considered nor evaluated at the same time. However, prerenal azotaemia also affects renal function and is not easily excluded at the first diagnosis when there is acute renal injury; the clinical cause of disease for each case is difficult to pinpoint exactly. Furthermore, any abnormality of another system would interfere with the renal system at the same time and thus cannot be excluded completely.
In conclusion, the present study showed that the use of RIFLE criteria is an appropriate method for classification of dogs with AKI, and this easy method can be used to evaluate AKI before more specific criteria are established for dogs. Furthermore, the new scoring system for dogs with AKI, developed using the multiple logistic regression approach, offers a better diagnostic performance for prognosis than RIFLE criteria alone. More studies are needed to support the use of RIFLE criteria in dogs.
The authors thank Professor T-H. Hsu for his kind permission to use the data from the Veterinary Medical Teaching Hospital, National Chung Hsing University.
Provenance not commissioned; externally peer reviewed
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.